Emotions for Everyone: A Low-Cost, High-Accuracy Method for Emotion Classification

Abstract The classification of emotions is of vital importance in health care, particularly in the context of early detection of cognitive disorders. Given the critical role of emotions as early indicators of cognitive health, this study addresses the need to develop effective and accessible classification methods. In this research, we present an innovative approach to emotion classification using a proprietary dataset and harnessing the power of deep learning. In particular, we use a specific, innovative combination of attentional layers and Long-Short Term Memory (LSTM) algorithms to achieve an emotion classification. A key differentiator of our methodology is the use of a compact and low-cost array of biometric sensors. This approach provides a cost-effective alternative to traditional systems, which often rely on more complex and expensive sensor arrays, such as those using electroencephalography (EEG). Despite the affordability of our sensor configuration, our classification model achieves an outstanding accuracy rate of 93.75%. This performance not only demonstrates the effectiveness of our method but also positions it at the forefront of emotion classification using these sensors. By significantly reducing cost while increasing classification accuracy, our method helps to push the boundaries of current state-of-the-art approaches and provides a novel and cost-effective solution for emotion classification and cognitive health monitoring.